{"id":5507,"date":"2024-09-30T06:15:00","date_gmt":"2024-09-29T21:15:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=5507"},"modified":"2024-09-30T06:15:00","modified_gmt":"2024-09-29T21:15:00","slug":"llama3-2-molmo-emova","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=5507","title":{"rendered":"Llama3.2, Molmo, EMOVA"},"content":{"rendered":"\n<p>\u5148\u9031\u306f\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u3067\u516c\u958b\u30e2\u30c7\u30eb\u3067\u3042\u308bLLM\u306e\u8a71\u984c\u304c\u591a\u304b\u3063\u305f\u3002Llama 3.2\u306fLlama\u306e\u30a2\u30c3\u30d7\u30c7\u30fc\u30c8\u3067\u3042\u308a90B\u3067GPT-4o mini\u306b\u5339\u6575\u3001Molmo\u306f72B\u3067GPT-4o\u306b\u7af6\u5408\u3059\u308b\u3068\u306e\u3053\u3068\u3002\u5546\u7528\u30e2\u30c7\u30eb\u306b\u516c\u958b\u30e2\u30c7\u30eb\u304c\u8ffd\u3044\u3064\u304d\u3064\u3064\u3042\u308b\u72b6\u6cc1\u3067\u4eca\u5f8c\u304c\u975e\u5e38\u306b\u697d\u3057\u307f\u3067\u3042\u308b\u3002<\/p>\n\n\n\n<p>\u516c\u958b\u30e2\u30c7\u30eb\u3067\u306f\u306a\u3044\u3088\u3046\u3060\u304c\u3001\u8907\u6570\u306e\u30e2\u30c7\u30eb\u3092\u7d44\u307f\u5408\u308f\u305b\u305fEMOVA\u306fGemini Pro 1.5\u3084GPT-4V\u4ee5\u4e0a\u3001GPT-4o\u306e\u30b9\u30b3\u30a2\u306e95%\u4ee5\u4e0a\u3092\u9054\u6210\u3068\u4e3b\u5f35\u3057\u3066\u3044\u308b\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Llama 3.2<br><a href=\"https:\/\/ai.meta.com\/blog\/llama-3-2-connect-2024-vision-edge-mobile-devices\/\">Llama 3.2: Revolutionizing edge AI and vision with open, customizable models (meta.com)<\/a><br><a href=\"https:\/\/huggingface.co\/collections\/meta-llama\/llama-32-66f448ffc8c32f949b04c8cf\">Llama 3.2 &#8211; a meta-llama Collection (huggingface.co)<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Multimodal Models\u00a0<\/strong>[146.2]<br>Molmo\u306f\u3001\u30aa\u30fc\u30d7\u30f3\u30cd\u30b9\u306e\u30af\u30e9\u30b9\u3067\u6700\u5148\u7aef\u306eVLM\u306e\u65b0\u305f\u306a\u30d5\u30a1\u30df\u30ea\u30fc\u3067\u3042\u308b\u3002 \u79c1\u305f\u3061\u306e\u91cd\u8981\u306a\u30a4\u30ce\u30d9\u30fc\u30b7\u30e7\u30f3\u306f\u3001\u4eba\u9593\u306e\u30a2\u30ce\u30c6\u30fc\u30b7\u30e7\u30f3\u304b\u3089\u53ce\u96c6\u3055\u308c\u305f\u3001\u65b0\u3057\u304f\u3066\u8a73\u7d30\u306a\u753b\u50cf\u30ad\u30e3\u30d7\u30b7\u30e7\u30f3\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3067\u3059\u3002 \u8fd1\u3044\u5c06\u6765\u3001\u30e2\u30c7\u30eb\u30a6\u30a7\u30a4\u30c8\u3001\u30ad\u30e3\u30d7\u30b7\u30e7\u30f3\u3001\u5fae\u8abf\u6574\u30c7\u30fc\u30bf\u3001\u30bd\u30fc\u30b9\u30b3\u30fc\u30c9\u3092\u3059\u3079\u3066\u30ea\u30ea\u30fc\u30b9\u3059\u308b\u4e88\u5b9a\u3067\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2409.17146v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2409.17146v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Wed, 25 Sep 2024 17:59:51 GMT)<\/li>\n\n\n\n<li>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u30b5\u30a4\u30c8\u306f<a href=\"https:\/\/molmo.allenai.org\/blog\">molmo.allenai.org\/blog<\/a>\u3001\u300cThe best-inclass 72B model within the Molmo family not only outperforms others in the class of open weight and data models but also compares favorably against proprietary systems like GPT-4o, Claude 3.5, and Gemini 1.5 on both academic benchmarks and human evaluation.\u300d\u3002PixMo (Pixels for Molmo)\u3068\u3044\u3046\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u69cb\u7bc9\u3001\u305d\u306e\u54c1\u8cea\u304c\u6027\u80fd\u5411\u4e0a\u306b\u5bc4\u4e0e\u3057\u3066\u3044\u308b\u3068\u306e\u3053\u3068\u3002<\/li>\n\n\n\n<li>\u30c7\u30e2\u306f<a href=\"https:\/\/molmo.allenai.org\/\">Molmo by Ai2 (allenai.org)<\/a>\u3001\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/huggingface.co\/collections\/allenai\/molmo-66f379e6fe3b8ef090a8ca19\">Molmo &#8211; a allenai Collection (huggingface.co)<\/a>\u3001Apache-2\u306eOSS\u3067\u3042\u308b\u3053\u3068\u3082\u51c4\u3044\u3002<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>EMOVA: Empowering Language Models to See, Hear and Speak with Vivid Emotions\u00a0<\/strong>[150.9]<br>GPT-4o\u306f\u3001\u591a\u69d8\u306a\u611f\u60c5\u3084\u58f0\u8abf\u3092\u6301\u3064\u58f0\u306e\u4f1a\u8a71\u3092\u53ef\u80fd\u306b\u3059\u308b\u30aa\u30e0\u30cb\u30e2\u30fc\u30c0\u30eb\u30e2\u30c7\u30eb\u3067\u3042\u308b\u3002 \u672c\u7814\u7a76\u3067\u306f,\u30a8\u30f3\u30c9\u30fb\u30c4\u30fc\u30fb\u30a8\u30f3\u30c9\u306e\u97f3\u58f0\u6a5f\u80fd\u3092\u5099\u3048\u305f\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb\u3092\u5b9f\u73fe\u3059\u308b\u305f\u3081\u306eEMOVA\u3092\u63d0\u6848\u3059\u308b\u3002 EMOVA\u306f\u3001\u8996\u899a\u8a00\u8a9e\u3068\u97f3\u58f0\u306e\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u306e\u4e21\u65b9\u3067\u6700\u5148\u7aef\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u521d\u3081\u3066\u9054\u6210\u3057\u305f\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2409.18042v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2409.18042v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 26 Sep 2024 16:44:02 GMT)<\/li>\n\n\n\n<li>\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u306a\u30e2\u30c7\u30eb\u3001\u300cEMOVA exceeds both GPT-4V and Gemini Pro 1.5 significantly on 10 out of 14 benchmarks, while for GPT-4o, EMOVA outperforms on both SEEDBench-Image and OCRBench, reaching over 95% of GPT-4o\u2019s performance on ALL evaluated benchmarks except RealWorldQA.\u300d\u3068\u306e\u3053\u3068\u3002LLaMA-3.1-8B\u00a0+InternViT-6B+ Speech\u30e2\u30c7\u30eb\uff08\u65e2\u5b58\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u3092\u30d9\u30fc\u30b9\u306b\u8457\u8005\u3089\u304cpre train\uff09\u306a\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u3002<\/li>\n\n\n\n<li>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u30b5\u30a4\u30c8\u306f<a href=\"https:\/\/emova-ollm.github.io\/\">EMOVA: Empowering Language Models to See, Hear and Speak with Vivid Emotion (emova-ollm.github.io)<\/a>\u3001<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>\u5148\u9031\u306f\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u3067\u516c\u958b\u30e2\u30c7\u30eb\u3067\u3042\u308bLLM\u306e\u8a71\u984c\u304c\u591a\u304b\u3063\u305f\u3002Llama 3.2\u306fLlama\u306e\u30a2\u30c3\u30d7\u30c7\u30fc\u30c8\u3067\u3042\u308a90B\u3067GPT-4o mini\u306b\u5339\u6575\u3001Molmo\u306f72B\u3067GPT-4o\u306b\u7af6\u5408\u3059\u308b\u3068\u306e\u3053\u3068\u3002\u5546\u7528\u30e2\u30c7\u30eb\u306b\u516c\u958b\u30e2 &hellip; <a href=\"https:\/\/devneko.jp\/wordpress\/?p=5507\" class=\"more-link\"><span class=\"screen-reader-text\">&#8220;Llama3.2, Molmo, EMOVA&#8221; \u306e<\/span>\u7d9a\u304d\u3092\u8aad\u3080<\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[223,251,293],"class_list":["post-5507","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-llm","tag-mllm","tag-oss"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5507","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=5507"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5507\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5507"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5507"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5507"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}